Got Questions? Get Answers.
Discover MakerZone

MATLAB and Simulink resources for Arduino, LEGO, and Raspberry Pi

Learn more

Discover what MATLAB® can do for your career.

Opportunities for recent engineering grads.

Apply Today

Thread Subject:
feature extraction.

Subject: feature extraction.

From: pankaj badoni

Date: 29 Oct, 2010 17:21:04

Message: 1 of 16

hi this one is my first post to this community...i am final year student and i am working on image processing, my problem is that i have to detect lesion in eye(Diabetic retinopathy).and for that i am using AM-FM models, so if any one can guide me how to approach the problem.

thanks.!!

Subject: feature extraction.

From: Sean

Date: 29 Oct, 2010 17:33:05

Message: 2 of 16

"pankaj badoni" <toni.badoni@gmail.com> wrote in message <iaevpv$a6j$1@fred.mathworks.com>...
> hi this one is my first post to this community...i am final year student and i am working on image processing, my problem is that i have to detect lesion in eye(Diabetic retinopathy).and for that i am using AM-FM models, so if any one can guide me how to approach the problem.
>
> thanks.!!

We can't guide you when we can't see your images. Post them somewhere like:
http://drop.io

Subject: feature extraction.

From: pankaj badoni

Date: 30 Oct, 2010 02:33:04

Message: 3 of 16

"Sean " <sean.dewolski@nospamplease.umit.maine.edu> wrote in message <iaf0gh$q7s$1@fred.mathworks.com>...
> "pankaj badoni" <toni.badoni@gmail.com> wrote in message <iaevpv$a6j$1@fred.mathworks.com>...
> > hi this one is my first post to this community...i am final year student and i am working on image processing, my problem is that i have to detect lesion in eye(Diabetic retinopathy).and for that i am using AM-FM models, so if any one can guide me how to approach the problem.
> >
> > thanks.!!
>
> We can't guide you when we can't see your images. Post them somewhere like:
> http://drop.io

It seems drop.io is not working now a days, can i put the link from where i got my dataset, the problem is like biomedical imaging and where features are non-linear and non-stationary, that's why i have opted for AM-FM models.

Subject: feature extraction.

From: ImageAnalyst

Date: 30 Oct, 2010 03:34:11

Message: 4 of 16

That's really terrible. Apparently Facebook bought drop.io and is
shutting down the best and easiest file sharing site out there. This
is very sad news. The ability to upload files without creating an
account made it very easy and quick to use. Hopefully some other web
site will take its place. Yet another reason to hate Facebook.

tinypic.com is not bad (http://tinypic.com/) - no account to create
but it requires the extra step of having to enter two captcha words,
and a then gives a list of URL's for your pictures that can be
confusing.

Failing that, just upload to any other web site you want. Preferably
one where we can see the image immediately without having to do any
additional clicks, or needing to create an account, and with the
ability to download the files easily.

I didn't understand your explanation of your problem - why you
couldn't give the URL to the original location of your data. What
does the data type and model type have to do with giving the URL?

Subject: feature extraction.

From: pankaj badoni

Date: 30 Oct, 2010 07:42:04

Message: 5 of 16

ImageAnalyst <imageanalyst@mailinator.com> wrote in message <9f64a51d-8183-4364-ac01-9d2c1e039579@s4g2000yql.googlegroups.com>...
> That's really terrible. Apparently Facebook bought drop.io and is
> shutting down the best and easiest file sharing site out there. This
> is very sad news. The ability to upload files without creating an
> account made it very easy and quick to use. Hopefully some other web
> site will take its place. Yet another reason to hate Facebook.
>
> tinypic.com is not bad (http://tinypic.com/) - no account to create
> but it requires the extra step of having to enter two captcha words,
> and a then gives a list of URL's for your pictures that can be
> confusing.
>
> Failing that, just upload to any other web site you want. Preferably
> one where we can see the image immediately without having to do any
> additional clicks, or needing to create an account, and with the
> ability to download the files easily.
>
> I didn't understand your explanation of your problem - why you
> couldn't give the URL to the original location of your data. What
> does the data type and model type have to do with giving the URL?

thanks a lot for replying...
this is the link of the dataset..
http://eyephoto.ophth.wisc.edu/ResearchAreas/Diabetes/DiabStds.htm

from here i am getting the images for my data, my just concern is to detect whether the image is lesion affected or not. when you will see the image, it has no major abnormality but their are few white spots, or the leaking vessels,or the dark spots, if i can extract these things then i can create a training set from these images and my work will be done, but since the problem(lesion) is not stationary and the lesions can occur at any position so here it is difficult to extract features.
i hope this time i have made my self clear.
   

Subject: feature extraction.

From: ImageAnalyst

Date: 30 Oct, 2010 13:51:57

Message: 6 of 16

Surely this must have been done before:
Have you checked the hundreds of papers on retinal image analysis
listed here:

http://iris.usc.edu/Vision-Notes/bibliography/medical853.html#Retinal%20Images,%20Analysis%20of%20Eye,%20etc.

and

http://iris.usc.edu/Vision-Notes/bibliography/medical854v1.html#Retinal%20Images,%20Angiography,%20Blood%20Vessels%20in%20the%20Eye

I bet at least one of them did exactly what you need to do.

Subject: feature extraction.

From: pankaj badoni

Date: 30 Oct, 2010 16:22:04

Message: 7 of 16

ImageAnalyst <imageanalyst@mailinator.com> wrote in message <0d4cab12-9d09-4660-b8d3-9380e040166d@g13g2000yqj.googlegroups.com>...
> Surely this must have been done before:
> Have you checked the hundreds of papers on retinal image analysis
> listed here:
>
> http://iris.usc.edu/Vision-Notes/bibliography/medical853.html#Retinal%20Images,%20Analysis%20of%20Eye,%20etc.
>
> and
>
> http://iris.usc.edu/Vision-Notes/bibliography/medical854v1.html#Retinal%20Images,%20Angiography,%20Blood%20Vessels%20in%20the%20Eye
>
> I bet at least one of them did exactly what you need to do.

thanks for replying... i have gone through many papers but i am not sure how to approach the problem with coding..don't know which filter would be working or which segmentation problem would be working.,,so bit messed up, so if you can help me in particular location where i can find code, so that i can try them on my image.

thanks.!

Subject: feature extraction.

From: ImageAnalyst

Date: 30 Oct, 2010 16:43:18

Message: 8 of 16

Which paper or algorithm are you trying to code up?

Can't your professor help you?

Subject: feature extraction.

From: pankaj badoni

Date: 31 Oct, 2010 03:04:04

Message: 9 of 16

ImageAnalyst <imageanalyst@mailinator.com> wrote in message <f3ab30b8-b5bd-408d-88a0-204b10ce6bf0@j25g2000yqa.googlegroups.com>...
> Which paper or algorithm are you trying to code up?
>
> Can't your professor help you?

well my prof is not going to help me, because the guy is having least knowledge of Image processing, but he is really cool.
my base paper is

Multiscale AM-FM Methods for Diabetic Retinopathy Lesion Detection.

so if anyone can tell how this thing works.

my next query is is their any format in images like "float"...because i have got one package which uses this type of format for images or rather i will give u link.
http://amfm.ou.edu/

in this they are using there own images whose format is "float", and i am having least knowledge of this thing, please just go through this one, may be you are deduce any thing.

thanks.!!

Subject: feature extraction.

From: ImageAnalyst

Date: 31 Oct, 2010 04:20:02

Message: 10 of 16

Sorry, but no one is going to read the paper and figure it out and
then tutor you on how to do it, although it does sound interesting.
You could do that as well as I or anyone else could. I can't even
find the time to read all the papers I want to read, much less the
ones that you want to read. It's good though that you've identified a
paper that does what you need to do. If you have the algorithm, even
in pseudocode, then we could help you with the syntax or efficiency of
your code.

float just means floating point. It's not any special kind of disk
file format. In MATLAB, it just means that the image is a floating
point array which will be of type single or double precision. For
most operations in MATLAB, the functions expect that floating point
images well be normalized to the range of 0-1.

Subject: feature extraction.

From: pankaj badoni

Date: 31 Oct, 2010 13:16:03

Message: 11 of 16

ImageAnalyst <imageanalyst@mailinator.com> wrote in message <219a878f-60ba-4f4f-b930-b92c7be11a1b@26g2000yqv.googlegroups.com>...
> Sorry, but no one is going to read the paper and figure it out and
> then tutor you on how to do it, although it does sound interesting.
> You could do that as well as I or anyone else could. I can't even
> find the time to read all the papers I want to read, much less the
> ones that you want to read. It's good though that you've identified a
> paper that does what you need to do. If you have the algorithm, even
> in pseudocode, then we could help you with the syntax or efficiency of
> your code.
>
> float just means floating point. It's not any special kind of disk
> file format. In MATLAB, it just means that the image is a floating
> point array which will be of type single or double precision. For
> most operations in MATLAB, the functions expect that floating point
> images well be normalized to the range of 0-1.

thanks...can u do small favor..just go through the images at this link, and then suggest me what is the format of the images.
http://amfm.ou.edu/
their is a link for images.

Subject: feature extraction.

From: ImageAnalyst

Date: 31 Oct, 2010 14:12:55

Message: 12 of 16

They don't appear to be any common format (I looked at the first two
bytes which identify formats in standard formats). They appear to be
a raw format, perhaps just a binary dump of the pixel values and maybe
or maybe not a header saying how many rows and columns. Their toolbox
must have a function that will read that format, but I didn't delve
into it that much. You could use their function to read them in and
write them out with imwrite() or save() into a more common format.

Subject: feature extraction.

From: pankaj badoni

Date: 1 Nov, 2010 03:18:03

Message: 13 of 16

ImageAnalyst <imageanalyst@mailinator.com> wrote in message <92d51104-1c6c-47e4-a863-21254b690b8a@26g2000yqv.googlegroups.com>...
> They don't appear to be any common format (I looked at the first two
> bytes which identify formats in standard formats). They appear to be
> a raw format, perhaps just a binary dump of the pixel values and maybe
> or maybe not a header saying how many rows and columns. Their toolbox
> must have a function that will read that format, but I didn't delve
> into it that much. You could use their function to read them in and
> write them out with imwrite() or save() into a more common format.


thanks alot.!!!
that was the thing i was asking for.
 one of my friend is having this problem he is having a mammogram image which is of 22 MB, and when he is operating in that image, the operation is taking a hell lot of time, so is this because of the hardware restrictions, or there is a need for code optimization.
platfrom os matlab.
another time i will be sending code also.

Subject: feature extraction.

From: pankaj badoni

Date: 1 Nov, 2010 03:21:03

Message: 14 of 16

ImageAnalyst <imageanalyst@mailinator.com> wrote in message <92d51104-1c6c-47e4-a863-21254b690b8a@26g2000yqv.googlegroups.com>...
> They don't appear to be any common format (I looked at the first two
> bytes which identify formats in standard formats). They appear to be
> a raw format, perhaps just a binary dump of the pixel values and maybe
> or maybe not a header saying how many rows and columns. Their toolbox
> must have a function that will read that format, but I didn't delve
> into it that much. You could use their function to read them in and
> write them out with imwrite() or save() into a more common format.

one thing more, if i want to convert my image to this format i.e of float, is their any technique by which we can do that thing.

Subject: feature extraction.

From: ImageAnalyst

Date: 1 Nov, 2010 03:30:18

Message: 15 of 16

On Oct 31, 11:21 pm, "pankaj badoni"
> one thing more, if i want to convert my image to this format i.e of float, is their any technique by which we can do that thing.
---------------------------------------------------------------------------
You can use double(), single(), or im2double() depending on what
precision you want and whether you want it to be nromalized to 0-1 or
not. Once you have a floating point array, you can look in that code
to figure out how to write it out to that .float extension file, if
you want it in that format. Like I said, it's not a standard format
but probably one of their own creation.

Subject: feature extraction.

From: Valeska Pearson

Date: 28 Mar, 2013 12:05:07

Message: 16 of 16

Hello

I'm a final engineering student this year in SA. So I saw you discussed the same topic which my project is on. Mine is specifically on automatic diabetic retinopathy detection. Could you give me tips in the starting of image processing in MATLAB. Or more specifically about image pre-processing which is the first stage.

Thanks for a reply

Tags for this Thread

What are tags?

A tag is like a keyword or category label associated with each thread. Tags make it easier for you to find threads of interest.

Anyone can tag a thread. Tags are public and visible to everyone.

Contact us